No — for most small business websites, adding an llms.txt file in 2026 is not worth prioritizing. The clearest evidence comes from an Ahrefs analysis of over 137,000 domains published this year: 28% of sites had already added an llms.txt file, but 97% of those files received zero requests in the month studied. The file has become a popular thing to add and an almost universally ignored thing once added.
That’s a strange enough finding that it’s worth unpacking properly, because “should we add llms.txt” is a question we now get from clients almost as often as “should we run Google Ads,” and the honest answer requires explaining what the file actually does versus what people assume it does.
What is llms.txt supposed to do?
llms.txt is a proposed standard — a plain-text file placed at yoursite.com/llms.txt — meant to give AI systems a curated, markdown-formatted summary of a site’s most important pages, similar in spirit to how sitemap.xml helps search engines and robots.txt tells crawlers what to avoid. The idea, when it was proposed, was appealing: instead of an AI model scraping and interpreting your whole site, you hand it a clean, pre-digested map of what matters.
The problem is that no major AI provider — not OpenAI, not Anthropic, not Google, not Perplexity — has ever confirmed that their systems actually fetch or use it in a standard way. It was proposed by one influential voice in the AI tooling space, adopted enthusiastically by the SEO and GEO community, and never formally adopted by the companies whose crawlers it was designed for.
What does the actual usage data show?
The Ahrefs study is the most rigorous public data we’ve seen on this, and the numbers are worth sitting with:
- 28% of analyzed domains published an llms.txt file — adoption clearly happened.
- 97% of those files got zero requests in the month studied. Only about 3% saw any traffic at all.
- Of the traffic that did arrive, 96% came from bots, but the majority of that bot traffic wasn’t even AI-related — it was SEO audit tools and general-purpose crawlers checking whether the file existed, not AI systems reading it for content.
- Genuine AI bot traffic made up only about 19.5% of the requests that did happen, split across AI agents, training crawlers, AI assistants, and retrieval bots — and coding-assistant tools showed up more than search-facing AI tools did.
- Perhaps the most important finding: no AI bot proactively went looking for a missing llms.txt file. They only fetch it when something explicitly points them there. It isn’t being discovered; it has to be handed over.
In plain terms: the file exists on over a quarter of sites analyzed, and almost none of those copies are doing anything.
So why did llms.txt become popular advice?
Because it’s an easy, concrete, low-cost action in a topic — AI search visibility — that otherwise feels foggy and hard to act on. “Add this file” is a satisfying answer to “how do we show up in ChatGPT,” even when the underlying mechanism hasn’t been validated. We understand the appeal; some of our own early GEO checklists included it before the adoption data caught up with the hype. It’s a good reminder that a tactic being widely recommended isn’t the same as a tactic being proven to work.
What should you do instead of chasing llms.txt?
The uncomfortable truth is that the things which actually move the needle for AI visibility are less novel and more work than dropping in one text file. Based on what we’ve seen work across client accounts and the broader GEO research:
1. Make your existing pages easy to extract, not just easy to crawl
AI systems that cite sources are pulling specific passages, not whole pages. Clear H2/H3 structure, a direct answer in the first two sentences of a section, and content that states facts plainly (rather than burying them in marketing language) gets extracted more reliably than a beautifully written page that makes a reader work for the point. We go deeper on this in our practical guide to getting cited in ChatGPT and Gemini answers.
2. Fix the boring technical basics first
Slow page loads, blocked resources, missing or broken schema markup, and pages that require JavaScript to render core content all reduce the odds of being crawled cleanly — by search engines and AI systems alike. If your site isn’t showing up in AI answers at all, this is usually a more productive place to look than llms.txt. We cover the specific fixes in why your site isn’t appearing in AI answers.
3. Earn the kind of mentions AI models already trust
AI answer engines lean heavily on third-party signals — reviews, comparison articles, forum threads, other sites citing you — the same way they lean on backlinks for traditional search. A curated llms.txt file can’t manufacture that trust; consistent, genuinely useful content over time can.
4. Track whether any of it is working
Before investing more time in AI visibility work, it’s worth establishing a baseline: are you already being cited anywhere, for what queries, and how often? Without that baseline, you can’t tell whether a change helped. We walk through why this matters in our breakdown of whether GEO is worth it in 2026.
Common questions about llms.txt
Will adding llms.txt hurt my SEO or GEO performance?
No. Adding the file does nothing harmful — it’s a small, static text file that sits quietly at your root domain. The risk isn’t harm, it’s opportunity cost: time spent formatting a perfect llms.txt is time not spent on the technical and content fixes that the data shows actually get read.
Should I remove llms.txt if I already have one?
No need. If it’s already published, leave it — there’s no penalty for having it, and on the small chance a tool does check for it, an outdated or missing file looks worse than a present one. Just don’t let it become a line item your team spends real hours maintaining.
Is llms.txt the same thing as robots.txt?
No, and mixing them up matters. robots.txt is a directive file that’s existed since the 1990s and is universally respected by search and AI crawlers — it tells bots what they’re allowed to access. llms.txt is a newer, unofficial proposal meant to summarize content for AI systems, and as the data above shows, it isn’t reliably fetched at all. Getting robots.txt right (not accidentally blocking AI crawlers you want visiting your site) matters far more than adding llms.txt.
What should I check first if I suspect AI crawlers can’t reach my site?
Start with your robots.txt file and confirm you’re not blocking known AI user-agents (GPTBot, ClaudeBot, PerplexityBot, and similar) either deliberately or by an old rule inherited from a previous developer. That single misconfiguration blocks far more AI visibility than a missing llms.txt file ever could.
Is there any case where llms.txt is still worth adding?
Yes, a narrow one. The Ahrefs data found that coding-assistant tools were among the more active fetchers of llms.txt files — which makes sense, since developers explicitly point tools like coding agents at a specific file when integrating with an API or SDK. If you run a developer-facing product, documentation site, or API, an llms.txt pointing to your docs is plausibly useful, because it’s being consumed by users who are directed to it on purpose, not passively discovered by a search-facing AI.
For most SME websites — local services, e-commerce, B2B without a developer audience — that use case doesn’t apply. Adding the file costs you almost nothing, so there’s no harm in having one, but it shouldn’t be where your GEO effort or budget goes. Spend that time on the extraction, technical, and trust-signal work above instead. It’s slower and less satisfying than dropping a text file in your root directory, but it’s the part that actually shows up in the data.